Engine Speed–Independent Acoustic Signature for Vehicles

Hüseyin Göksu

doi:10.1177/0020294018769080

Abstract

A vehicle, when running, makes a complex sound emission from the engine, the exhaust, the air conditioner, and other mechanical parts. Analysis of this sound for the purpose of vehicle identification is an interesting practice which has security- and transportation-related applications. Engine speed variation, which causes shifts in the frequency content of the emissions, makes Fourier-based methods ineffective in terms of providing a stable signature for the vehicle. We search for an engine speed–independent acoustic signature for the vehicle, and for this purpose, we propose wavelet packet analysis rather than traditional time- or frequency-domain methods. Wavelet packet analysis, by providing arbitrary time–frequency resolution, enables analyzing signals of stationary and non-stationary nature. It has better time representation than Fourier analysis and better high-frequency resolution than wavelet analysis. Under varying engine speed, sound emissions are recorded from four cars and analyzed by wavelet packet analysis. Wavelet packet analysis subimages are further analyzed to obtain feature vectors in the form of log energy entropy, norm entropy, and energy. These feature vectors are fed into a classifier, multilayer perceptron, for evaluation. While norm entropy achieves a classification rate of 100%, log energy entropy and energy achieves classification rates of 99.26% and 97.79%, respectively. These results indicate that, wavelet packet analysis along with norm entropy and multilayer perceptron provides an accurate vehicle-specific acoustic signature independent of the engine speed.

Highlights

A vehicle, when running, produces a complex sound emission from its engine, the exhaust, air conditioner, and other mechanical parts and, if it is moving, from tires and air friction
In search for an engine speed–independent acoustic signature for vehicles, we propose a method based on Wavelet packet analysis (WPA) of the sound signals for the aforementioned reasons
Reduction, which might reduce the complexity of the method, can be considered as a future work, but it is not essential since we have reached 100% classification accuracy, so we have found the right feature which provides a precise engine speed–independent acoustic signature for vehicles: WPA followed by norm entropy and multilayer perceptron (MLP)

Summary

Introduction

A vehicle, when running, produces a complex sound emission from its engine, the exhaust, air conditioner, and other mechanical parts and, if it is moving, from tires and air friction. The possibility of identification and diagnosis of vehicles based on the analysis of sound emissions has been investigated by researchers. Is a chronological review of time-domain, frequency-domain, time–frequency domain, and hybrid analysis methods used in vehicle acoustic signal analysis. Some of the researchers used time-domain features, for example, Mazarakis and Avaritsiotis[1] used a time-domain encoding and feature extraction to classify tracked vehicles and heavy trucks using their acoustic and seismic signatures; Paulraj et al.[2] used multi-frame time-domain features for classification of moving vehicles; Wang and Zhou[3] used improved time-encoded signal-processing algorithm for feature extraction in acoustic vehicle-type recognition; Rahim et al.[4] used time-domain features for the classification of moving vehicles; Paulraj et al.[5] used autoregressive modeling for vehicle-type classification; Mayvan et al.[6] used quadratic discriminant analysis to classify audio signals of passing vehicles to bus, car, motor, and truck categories based on features such as short-time energy, average zero-crossing rate, and pitch frequency of periodic segments of signals; and Ishida et al.[7] used dynamic time warping to design an acoustic vehicle-count system

Methods

Results

Conclusion